Extraction Of Sandy Land Information Based On Object-Oriented Method From Remote Sensing Data

Sandy lands are currently facing the most serious ecological problems in China. With thedeveloping society and the increasing population, our country is facing growing pressure onthe environment. Sandy lands not only led to the sharp drop in the available land resources,deterioration of ecological environment, the guide also caused huge economic losses to deepenthe extent of poverty of the sand people in the region. To this end, to keep abreast of the statusquo and development trend of sandy lands, it is of great significance for the development ofland desertification prevention and control strategies.Access to information of sandy lands, the development of remote sensing technologyprovides a new means of technology, especially for fast, accurate access to a wide range ofsandy lands. Conventional observing mode of the remote sensing image data cannot beanywhere in the material environment different to adjust adjust the sensor integration time andexposure parameters, thereby limiting the potential applications of remote sensing data.Intelligent observation technology of small satellites is nearly decade years of development ofthe Advanced Land Observing techniques, its observation mode by surface features intelligentrecognition of the environment and thus intelligent adjustment, greatly enhances the potentialof small satellite data in the remote sensing to monitor the application of resources and theenvironment. Due to high reflectivity of the arid areas, arid areas vegetation easily beconcealed and the traditional mode of satellite observations is not conducive to the extractionof the large surface vegetation information, the development of advanced intelligentobservation technology for small satellites, the extraction of sand vegetation complexinformation may be provided.Taking Otindag sandy land and its surrounding areas as study area, study was carried outon sandy information extraction technology research using intelligent image of the Beijing-1small satellite data sources based on the object-oriented and explore the intelligent dataextraction of the sandy lands information potential applications to summarize and mastery of the sandy lands based on object-oriented information extraction technology and at the sametime provide a more effective means for the desertification of land information extraction and awide range of sandy lands monitoring.. The main results and conclusion of this dissertation areas follows:1. Object-oriented methods and maximum likelihood method are carry out to extract theinformation of study area. The results show that object-oriented extraction of desertificationland information, the overall accuracy comes to83.80%and the Kappa coefficient is0.76;using the maximum likelihood method the overall accuracy reached75.70%, and the Kappacoefficient is0.65.2. In the object-oriented process of extracting sandy land, using four different scales layerto extract the characteristics of different objects types, through the establishment of surfacefeatures multiple feature map to select the optimal feature a variety of features of the targetobjects(Green Ratio, Ratio Red,Mean nir,NDVI and so on) to extract the target objects, andthe establishment of a classification decision tree to extract the sandy land information makesthe classification more reasonable, resulting in higher classification accuracy.3. Though BJ-1intelligent data, conventional data and Landsat-5TM image data sandyland information extraction results accuracy evaluation of contrast, the results show that: BJ-1intelligent data extraction desertification overall accuracy is85.55%, Kappa coefficient of0.80,the overall accuracy of the BJ-1conventional data to extract information is81.39%and Kappacoefficient is0.73; the overall accuracy of the Landsat-5TM data to extract information is90.81%and Kappa coefficient is0.86. It indicated that Landsat-5TM extraction of sandy landinformation is the best, BJ-1intelligent observation data followed and BJ-1conventionalobservational data is poor.4. From the water and saline extract of the three data mapping information extractionaccuracy was97.06%and user accuracy of100.00%, three data are consistent; desertificationinformation extraction, the BJ-1intelligent data flow of sandsemi-fixed sand dune and fixedsand dune, respectively78.21%,68.63%,89.38%; the Landsat-5TM data was81.37%,79.77%,93.76%; BJ-1routine observation data was69.11%,58.62%,82.88%, we can see the BJ-1 intelligent data desertification information extraction than conventional data of BJ-1has anadvantage, especially in the flow of sand and semi-fixed sandy land; compared with theLandsat-5TM images, the BJ-1intelligent dataextraction of semi-fixed and fixed sandy landinformation extraction gap, but extract the flow of sand, the results of two kinds of dataextraction accuracy less.